Better Buy: SK Hynix vs. Micron

The near-term winner is not just SKHY; it is any portfolio that owns the supply-constrained part of the AI stack while avoiding the pieces that become substitutable once capacity normalizes. HBM remains the bottleneck for accelerator shipments, so a leader with a privileged allocation relationship should keep pricing power longer than the broader DRAM market; that is constructive for NVDA in the near term because it reduces component risk, but the bigger economic transfer accrues to the memory vendor until supply catches up.

The market is likely underestimating how quickly this can swing from scarcity to oversupply once capex finally lands. Historically, memory equities peak on peak margins, not peak demand, and the first sign of a capacity inflection will be multiple compression before revenue rolls over. That makes MU the higher-beta way to express the cycle, while SKHY has the cleaner share story but also the greatest exposure if rivals or customers push qualification of second sources.

The contrarian point is that the new U.S. listing may drive a temporary valuation catch-up without changing intrinsic economics: liquidity and accessibility can tighten the spread for a few weeks, but they do not remove cyclicality. The more durable edge is execution on next-generation HBM nodes and packaging yields. If SKHY’s lead narrows or if capex announcements from peers suggest supply is arriving faster than demand, the thesis should be treated as a 6-12 month fade rather than a buy-and-hold compounder.